Recognising tones by tracking movements – How infants may develop tonal categories from adult speech input

Previous research has shown that the perception of speech in infants moves gradually from being language-general to being language-specific during the first year of life. Recently, it was found that infants learning a tone language begin to show particular response patterns to tones in their native language during their first year [1]. In this study we explore the relevance of tone production theory in understanding how the learning and the perception of Mandarin tones could have been accomplished despite highly variable speech input. The Target Approximation model [2] predicts that it is possible to infer underlying pitch targets from the manners of F0 movements, for they may directly reflect the characteristics of intended goals. Using the production data of multiple speakers in connected speech from [3], we trained a self-organising neural network with both F0 profiles and F0 velocity profiles as input. The network’s performance indicates that velocity profile distribution in the network formed distinct regions of clustering neighbourhoods representing each tone. The finding points to one way through which infants can successfully derive at phonetic categories from adult speech, namely, by extracting underlying phonetic targets based on information directly reflecting production.